Processors - Page 5

A new report via VideoCardz, citing Intel's internal Nova Lake-S roadmaps as the source, has lifted the lid on specs and details for the company's upcoming next-gen Intel Core Ultra 400 Series desktop processors. According to the materials obtained, the new CPUs will be tile-based like Arrow Lake-S and will introduce lower-power E-cores called LPE-cores.

They'll also support the new LGA1954 socket but will retain the same cooler compatibility as the report, noting that the plan is for this socket to be forward-compatible with upcoming Intel CPU hardware, similar to how AMD's AM5 and AM4 platforms supported multiple generations of chips. The new Core Ultra 400 Series will also include dual-die processors, paving the way for higher core counts.

And with that, what makes this report interesting is that it offers what could become the entire lineup, from the new Core Ultra 3 to the new Core Ultra 9. This includes everything from a 6-core CPU to a 175W 52-core model, which may or may not be for the consumer market. That said, we'd love to see Intel release a Threadripper-style 52-core Core X CPU for enthusiasts.

AMD's best CPU lineup is set to get a new entry on April 22nd. The Ryzen 9 9950X3D2 is the world's first CPU to feature dual 3D V-Cache, with a combined cache of 208MB including L2. However, that premium specification comes with a premium price tag, and its eye-watering $899 asking price for a consumer processor is pushing many gamers away.

Lucky for you, another top gaming CPU in the X3D lineup is currently available for almost half that price. The Ryzen 7 9800X3D is on sale for 14% off on Amazon right now, bringing it to its lowest price yet on the retailer. The discount brings it down to $409.95 from the usual $479, cutting roughly $70 off the original MSRP. The previous all-time low was $419.95 in late March, so this latest deal pushes it another $10 below that record.

The new price positions the Ryzen 7 9800X3D as an even better value than the slightly refreshed Ryzen 7 9850X3D, which currently costs $489. With only a 2-3% performance difference, most users will see virtually identical real-world gaming performance, making the 9800X3D's $80 lower price a compelling advantage for gamers seeking the best value per dollar.

Over the weekend, retailer listings started popping up suggesting AMD's upcoming Ryzen 9 9950X3D2 could cost close to $1,000. AMD has now lifted the curtain on pricing. The CPU will retail at $899, confirmed today on X by AMD's David McAfee, VP and GM of Ryzen CPU and Radeon Graphics. The new flagship desktop CPU will be available on April 22.

Now for the comparison. The $899 price tag is at least lower than what early listings suggested, but it still represents a $200 premium over the regular Ryzen 9 9950X3D, which launched at $699. That works out to around a 29% increase. With the 9950X3D sitting at $675 on Amazon right now, the gap at launch is closer to 33%.

That $200 premium goes toward making the 9950X3D2 the first dual-cache X3D CPU from AMD, applying 3D V-Cache packaging to both eight-core CCDs rather than just one. Worth noting is that AMD is not positioning this as a gaming chip. While X3D processors like the Ryzen 7 9850X3D are aimed at gamers chasing high frame rates, AMD is pitching the 9950X3D2 as a workstation-level CPU "built for developers and content creators tackling complex workloads and massive datasets."

Overclocking wizard and PC enthusiast Der8auer delidded Intel's latest Core Ultra 7 270K Plus to see how far he could push the CPU. He found his specific sample was capable of achieving 5.8GHz on the P-cores and 5.3GHz on the E-cores after delidding the chip. With this configuration, he was able to achieve a Cinebench R23 score 11% better than stock speed and outperform his own previous R23 runs with a delidded and overclocked Core Ultra 9 285K.

Der8auer showed his overclocking process and how to delid the 270K Plus with special tools. The YouTuber revealed that delidding LGA 1851 CPUs is more difficult compared to AMD Ryzen and previous-generation Intel CPUs. Delidding the CPU the usual way by prying the IHS away from the die could damage the chip.

To remedy this, Der8auer built a heating solution when the Core Ultra 9 285K first came out to address this problem on Arrow Lake-S processors. The heating process involves running the chip up to almost 170C to melt the indium metal connecting the die to the IHS, moving the IHS in one direction, then heating the CPU a second time to remove the IHS. Der8auer successfully used this method to delid his Core Ultra 7 270K Plus.

Someone at Intel seems to be hard at work updating the product pages recently, and in the process, announcing new CPUs. We recently reported on the Core Ultra 7 251HX, a mobile chip in the Arrow Lake-HX lineup, which was quietly made official when its product page went live without any announcement. Now, another CPU is following the same trend, this time from the Panther Lake lineup.

The Core Ultra X9 378H is a new CPU that is part of Intel's high-end Core Ultra 300 lineup. It features a 16-core/16-thread configuration with 4 P-cores, 8 E-cores, and 4 LPE cores. Its P-core boost clock is 5.0 GHz, while the E-cores can boost up to 3.8 GHz. The chip's base TDP is 25W, with a maximum turbo power rating of 80W. None of these specs are exactly groundbreaking, but the Core Ultra X9 378H still has its place in the lineup for one reason.

On closer inspection, it's apparent that the Core Ultra X9 378H is identical to the already existing Core Ultra X7 368H. Both CPUs have the same 16-core/16-thread layouts, and even the underlying core clocks, TDP, and iGPU specs are the same. It even has the same 18MB of L3 cache and supports LPDDR5X at 9600 MT/s. So why did Intel bother releasing the same CPU with just a different name?

It looks like Intel was tired of the leaks and speculations surrounding its "mystery" Arrow Lake-HX CPU. As we recently reported, the Core Ultra 7 251HX was spotted on listings for laptops from vendors MSI and Lenovo. Now, only a few days after the report, Intel has silently made the Core Ultra 7 251HX official without any announcement.

The Core Ultra 7 251HX is part of the Arrow Lake lineup, and slots in just under the Core Ultra 7 255HX. Instead of having 20 cores and threads like its sibling, the 251HX has 18 cores and 18 threads, of which 6 are performance-focused P-cores, and 12 are E-cores. With this core configuration, the Core Ultra 7 251HX sits between the 14-core Ultra 5 245HX and the 20-core Ultra 7 255HX.

The aforementioned laptop listings also had some unconfirmed specs that have now been made official. As mentioned in the listings, the Core Ultra 7 251HX indeed has a 2.9 GHz P-core base clock and a 55W TDP. Of course, the maximum turbo power can reach up to 160W, depending on the type of workload and available thermal headroom.

A new update from hardware sleuth Jaykihn on X (formerly Twitter) suggests that Intel's next-generation Nova Lake family will not feature the Xe4 "Druid" architecture after all. While earlier rumors pointed to a hybrid Xe3/Xe4 setup, it appears the platform is sticking entirely to the Xe3 family.

Nova Lake is set to be Intel's next major architecture launching across the desktop and mobile segments of the market next year. The lineup is expected to debut Coyote Cove P-cores and Arctic Wolf E-cores, which should bring about a sizeable boost in IPC (Instructions Per Clock Cycle). Nova Lake reportedly marks a return to a Conroe (Core 2 Duo) style shared L2 cache hierarchy for its P-cores. This, along with the use of a dual Compute Tile layout, allows configurations to reach 52 cores on desktop and 28 on mobile, paired with specialized bLLC (Big Last Level Cache) to tackle AMD's X3D family.

The update suggests Nova Lake's integrated GPU will use the Xe3-LPG architecture, similar to Panther Lake's B390. This is technically classified as a Battlemage GPU, though with some small refinements.

Intel launched the Core Ultra 200S Arrow Lake CPUs in 2024, but they didn't see the success the Blue Team had hoped for. Then came the newer Core Ultra 200S Plus "Arrow Lake Refresh" lineup, which looked more competitive at launch but has recently seen price hikes. As a result, many users are turning back to Intel's Raptor Lake CPUs, with the 13th and 14th-Gen chips still offering strong multithreading and gaming performance even by today's standards. However, finding a CPU like the Core i5-14600K at a reasonable price is not easy right now, and that is exactly what Intel is looking to fix.

In light of the ongoing memory shortage and rising demand for DDR4-compatible CPUs, Robert Hallock, VP and GM of the Enthusiast Channel at Intel, told Club386 that Raptor Lake is "an extremely fast product and is not going away any time soon." He added that Raptor Lake is a "big part of our strategy, and will continue to be abundantly available."

This is welcome news for those who couldn't make the jump to DDR5 in time, as Intel is not closing the door on its DDR4-compatible LGA1700 platform. Raptor Lake desktop chips on LGA1700 give builders the choice between DDR4-3200 and DDR5-5600, keeping them relevant for anyone who wants to reuse existing memory rather than move to a more expensive DDR5-only platform.

Intel's Wildcat Lake CPU lineup has been a hot subject in the news lately. They were first spotted on Crossmark and Geekbench, giving us a rough idea of their performance levels. Then, another leak outlined the total number of SKUs and their specifications. Now, even more information has surfaced from OEM Advantech, which has posted a technical datasheet that mentions the specs of various Wildcat Lake CPUs.

According to the datasheet for the MIO-5356, three yet unreleased processors are supported on the platform: the Core 7 350, the Core 5 320, and the Core 3 305. Interestingly, the datasheet also lists detailed specs of the CPUs. The documentation clearly states that the model supports "Next Gen" Intel Core processors.

The Core 7 350 apparently has a maximum frequency of 4.8 GHz and a base clock of 1.5 GHz. It uses the same 2P+0E+4LPE core configuration that was previously leaked, and has a TDP of 15W. As a reminder, Wildcat Lake CPUs omit traditional "Darkmont" E-cores in favor of LPE cores.

AMD recently revealed its latest entry in the Ryzen 9000 lineup, the Ryzen 9 9950X3D2 Dual Edition, the world's first CPU with a dual 3D V-Cache design. While the Red Team shared detailed specifications, including core count, total cache, TDP, and an April 22 launch window, it did not reveal pricing. Now, with only a few weeks left until launch, early retail listings have started to surface, giving us our first glimpse at possible pricing.

The listings, spotted by Wccftech via @momomo_us, appeared on multiple UK and Canadian retailers showing the Ryzen 9 9950X3D2 Dual Edition listed at $958 and $1,196, respectively.

Popular Canadian retailer PC-Canada.com listed the 9950X3D2 Dual Edition at CAD $1,373, which translates to around US$984. That puts it almost $300 above the launch price of its single-V-Cache sibling, the Ryzen 9 9950X3D ($699 at launch). Another Canadian store, ShopRBC, listed the new CPU at CAD $1,375, which comes to around US$986.

The bad news just keeps coming for DIY PC builders and enthusiasts, as Intel is reportedly preparing yet another price increase for its CPUs. Earlier, we reported that Intel would implement a 10% price increase by the end of March, and now it seems like that won't be enough for Intel.

Chinese market research firm Minutes Logic Society reports that Intel plans to add another price increase on top of the price hikes already implemented in February and March. The new price increase will reportedly take effect in May and affect Intel's entire CPU portfolio. It seems the total cumulative goal for the price hike is 30% above 2025 prices.

The main driving force behind these price increases is the raging demand from AI data centers and large machine-learning enterprises. We have already seen the DRAM market ravaged by a global memory shortage that is expected to last until 2030. Previously, AMD also followed Intel's lead and planned to increase its CPU prices by 15%. It seems like the pricing snowball is barreling down the slope with no end in sight.

A new leak from renowned Intel tipster Jaykihn claims the premium Nova Lake-S desktop SKU will jump from 42 to 44 cores. This shift has left Intel with many 6P+12E tiles. These tiles, which feature bLLC (Big Last Level Cache) integrated directly on the chip, are now said to be repurposed for a new segment of mainstream 'non-K' gaming CPUs, set to rival AMD's next-generation X3D chips.

Leaks from last year detailed four primary bLLC configurations for Intel: a 52-core and 42-core flagship (both utilizing dual-compute tiles), alongside 28-core and 24-core single-tile variants. For the unaware, bLLC is Intel's response to AMD's X3D technology. While AMD vertically stacks a separate cache chiplet on top of the CPU, Intel integrates the bLLC directly into the Compute Tile's silicon, which will add to manufacturing costs.

According to Jaykihn, Intel is shifting its high-end Nova Lake-S configuration from a 42-core to a 44-core design. This change removes the asymmetrical configuration planned for the original model. That model would have required pairing a fully-enabled 8P+12E compute tile with a binned 6P+12E tile, both with bLLC. The updated 44-core model likely uses two 8P+12E tiles. That being said, the leftover 6P+12E tiles, which still have 144MB of bLLC, are now being reallocated to a new segment of mainstream 'non-K' gaming CPUs. These CPUs are expected to feature 4 LPE cores in the SoC Tile.

A new milestone was reached in the hardware modding community, as folks over on Overclock.net have successfully booted the Core 9 273PQE "Bartlett Lake" CPU to Windows. This is a monumental achievement for the modding community, as the CPU was never designed for consumer motherboards and is marketed exclusively by Intel for embedded applications and edge devices.

The latest update was part of a continuous modding effort by community member "kryptonfly," who has been trying to get the Core 9 273PQE to boot to Windows for a while now. On the overclock.net forum, the modder posted about a new milestone: he had rewritten the BIOS of a consumer Z790 motherboard to get the CPU to boot into Windows.

According to the modder, he used an ASUS Z790-AYW OC WiFi motherboard with a BIOS rewritten with AI assistance, and missing support components were then added. The board could then detect the Core 9 273PQE, but a successful boot sequence could not be achieved. Now, with a few additional steps to fool the boot sequence using a Raptor Lake spoof, the system can finally boot into Windows.

A new leak from hardware sleuth Jaykihn on X (formerly Twitter) suggests that Intel's next-generation Nova Lake HX family of processors will top out at a 28-core configuration. These CPUs are set to replace the incumbent Arrow Lake HX series, which is scheduled to arrive by late 2026 or early 2027.

Following the architectural shifts seen in Meteor Lake through Panther Lake, Intel's Nova Lake is expected to further refine this modular approach. At the microarchitectural level, Nova Lake is expected to pair "Coyote Cove" P-cores with "Darkmont" E-cores. While Darkmont has debuted with Panther Lake, details on the specific IPC gains and structural changes in Coyote Cove remain unknown.

The flagship desktop configuration is rumored to be a massive 3D package with several tiles, with up to 52 cores (16P+32E+4LPE), 48 of which are split across two Compute Tiles (each housing up to 24 cores) expected to be fabbed on Intel 18A or TSMC's N2 node, a centralized SoC Tile (expected to house the 4 LPE cores), a GPU Tile, and potentially a dedicated I/O tile.

Information about Intel's Wildcat Lake has been slowly emerging over the past few weeks. CPUs belonging to this yet unofficial lineup were first spotted on Geekbench and CrossMark, with another Core 3 304 entry surfacing later. Now, more information about Intel's Core 300 "Wildcat Lake" lineup has been uncovered by reputable leaker @Jaykihn on X (formerly Twitter).

According to the leak, the Intel Core 300 lineup will feature 6 SKUs, ranging from the base Core 3 304 to the flagship Core 7 360 CPU, with three Intel Core 5 SKUs in between. These CPUs are designed for efficiency, since they are mobile chips, which is why the leaked TDP ratings range from 15W to 35W across all SKUs.

Each CPU features a hybrid configuration, with two "Cougar Cove" P-cores paired with four low-power efficiency (LPE) cores. This configuration completely discards the "Darkmont" E-cores in favor of the LPE cores.

Intel recently launched its Arrow Lake refresh CPUs in the form of the new Intel Core Ultra 7 270K Plus and Intel Core Ultra 5 250K Plus, both of which we recently reviewed. When it comes to the more affordable Core Ultra 5 250K Plus, we praised its multi-threaded performance and value for money, with the $199 USD price point feeling spot-on in a consumer technology market where prices are seemingly increasing everywhere.

Part of the value also derives from the integrated Arc or Xe graphics that offer impressive iGPU performance for a desktop processor. However, Intel is planning to debut a 'KF' Core Ultra 5 250KF Plus variant on April 3, which will reportedly keep the same specs and optimizations but disable the iGPU - making it a pure CPU that requires a dedicated discrete GPU for a PC to boot.

Now, when it comes to desktop PCs, discrete GPUs - especially when it comes to gaming - are essential, and many systems out there rarely, if ever, make use of a CPU's iGPU. Of course, when it comes to troubleshooting hardware or GPU issues, having an iGPU can be a lifesaver, but the new Core Ultra 5 250KF Plus looks to be an option for those looking for an affordable CPU that lets them save some money or put that toward other components.

A new update from Geekbench sheds light on how Intel's new iBOT (Intel Binary Optimization Tool) works under the hood. The team conducted a series of tests on a Panther Lake-equipped laptop and found that it uses checksum-based matching to run verified apps. Most notably, the analysis suggests the tool goes beyond simple instruction reordering. Instead, it replaces unoptimized instruction sequences with optimized alternatives, for example, vectorized code.

iBOT was announced alongside Intel's debut of its Core Ultra 200 Plus processor lineup. Intel benchmarks or profiles select applications in its labs to identify unoptimized instruction sequences. If it finds instructions that can be optimized to better fit the architecture, it restructures and reorganizes them. Intel called this analogous to Tetris. The company explicitly stated that iBOT does not skip any instructions and that no reverse engineering is involved, as the optimization is performed post-link while the tool runs with user-mode privileges.

One of the supported applications for this utility was surprisingly Geekbench, which, on average, increased scores by 8%. Due to a lack of official technical details, Geekbench flagged all benchmarks conducted on CPUs that supported iBOT as potentially invalid. Following a week of rigorous internal testing, the developers have implemented a more precise detection method that isolates iBOT-optimized runs from standard ones.

Samsung's foundry division has set a goal of introducing 1nm process technology by 2030, which the company refers to as the 'dream semiconductor' process. According to a new report from Business Korea, citing industry sources, Samsung's current plan is to complete research and development of its 1nm semiconductor process by 2030, after which it will move to mass production.

To achieve 1nm, Samsung is set to deploy new 'fork sheet' technology that will allow it to effectively halve the size of its cutting-edge 2nm process, which is manufactured using Gate-All-Around (GAA) technology. According to the report, the 'fork sheet' minimizes the distance between GAA elements in a way that's similar to inserting a fork to free up space and allow for more transistors to be packed in the same area.

One analogy used in the report is that it would be like removing lawns from a housing complex and using the space to add more housing. Samsung Electronics' foundry business is second only to TSMC, which holds an almost 70% market share. Naturally, TSMC is also pursuing 1nm within a similar timeframe, with the report indicating it is set to adopt 'fork sheet' technology to achieve it.

One of the key features that made AM4 so popular and successful was its longevity. AMD had committed to supporting the platform through multiple Ryzen generations, and they ultimately stuck to it, which was an admirable decision. Users who started using the platform with first-generation Ryzen 1000-series CPUs can just drop in a Ryzen 9 5950X and get a massive upgrade with little more than a BIOS update.

While the current AMD Zen 5 platform isn't exactly "old", it has already seen a couple of generations of Ryzen CPUs. The Ryzen 7000 series "Zen 4" and Ryzen 9000 series "Zen 5" chips both support the AMD AM5 platform and its associated chipsets. Now, AMD is doubling down on its promise of longevity and giving its customers more confidence that the AM5 socket will be here to stay through 2027 and beyond.

Speaking in the launch video of the AMD Ryzen 9 9950X3D2 Dual Edition CPU, AMD Senior Vice President and General Manager of Computing and Graphics Jack Huynh gave this statement.

A series of new next-gen Zen 6 server EPYC processors from AMD have debuted on the open-source benchmarking platform OpenBenchmarking, thanks to Olrak29_ at X (formerly Twitter). These appear to be the first public sightings of the upcoming silicon, though performance is not indicative of the final result, as we're still dealing with early silicon. Codenamed 'Venice', AMD is on track to launch these 6th Gen EPYC processors in the second half of 2026.

Last April, AMD officially unveiled the codename and some architectural details for its next-generation EPYC family. The EPYC 9006 family is designed with the Zen 6 architecture and fabricated using TSMC's 2nm process. Current leaks underpin improved core counts, greater cache, and enhanced I/O capabilities, thanks to the new design and the higher density offered by the updated process node.

If we're to believe leaks, each Zen 6 CCD is moving from the standard 8-core structure to 12 cores, with 48MB of L3 cache, up from 32MB. The dense Zen 6c variant sees an even bigger jump from 16 cores to 32 cores per CCD, accompanied by 128MB of L3 cache (up from 32MB). This means that for a fully decked-out chip with eight Zen 6c CCDs, we're looking at 256 cores and 1GB of L3 cache. While industry insiders have teased this flagship spec, current data suggests the standard Zen 6 counterpart may top out at 96 cores with a similar eight CCD layout, unlike sixteen CCDs with Turin.